Identifying transcriptomic predictors of brodalumab response in psoriasis using CART analysis

Abstract

Precision medicine is a topic of growing interest in psoriasis. Many novel biologics are now available to clinicians and identifying who will be a responder or non-responder to a given biologic prior to treatment is an exciting area of inquiry with strong potential clinical utility. In the present study, we use an interpretable classification and regression tree (CART) model to predict week 12 PASI75 and PASI90 response to brodalumab treatment based on clinical variables and transcriptomic data from lesional biopsy tissue samples. We identify KRT16 normalized RNA expression levels and BMI as pre-treatment predictors of a PASI75 response and FERMT1, HLA_DQA1, TMPRSS11D, and S100P normalized RNA expression levels as pre-treatment predictors of a PASI90 response. The CART models demonstrated strong AUC values for the PASI75 (0.90) and PASI90 (0.88) analyses. Taken together, focused transcriptomics has the potential to be used clinically for the pre-treatment prediction of treatment response.